https://github.com/banyc/sharpnetwork

A bare-metal implementation of neural networks. Define networks in Numpy.NET. More than encapsulations.
https://github.com/banyc/sharpnetwork

csharp educational from-scratch neural-network numpy oop

Last synced: about 3 hours ago
JSON representation

A bare-metal implementation of neural networks. Define networks in Numpy.NET. More than encapsulations.

• Host: GitHub
• URL: https://github.com/banyc/sharpnetwork
• Owner: Banyc
• Created: 2020-10-29T17:56:40.000Z (about 4 years ago)
• Default Branch: master
• Last Pushed: 2020-11-24T11:58:12.000Z (almost 4 years ago)
• Last Synced: 2023-10-20T18:07:26.092Z (about 1 year ago)
• Topics: csharp, educational, from-scratch, neural-network, numpy, oop
• Language: C#
• Homepage:
• Size: 955 KB
• Stars: 4
• Watchers: 3
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# NumSharpNetwork

A deep dive into neural networks.

Defining network in Numpy.NET. For more information, go to [this repository](https://github.com/zllz4/numpy-network)

## Screenshot

Convolution on MNIST:

![](img/2020-11-10-22-12-00.png)

## Purpose

- For education rather than for production.

- OOP-friendly and maintainable machine learning code.

- Reduce stupid errors before compile time.

- Prefer robust coding than less typing.

## Restriction

Currently Windows only.

## Run

Suppose the `pwd` is the root folder of the project.

```bash

cd src/NumSharpNetwork.Client

dotnet restore

dotnet run

```

## Notices

- `lossXXXXXGradient` indicates `d_loss / d_XXXXX`

- `lossXXXXXGradient.shape == XXXXX.shape` 

## Credit

- The origin/inspiring project - 

- A good source of examples - 

## Todo

- [ ] Add test cases for each layer